arm_cmplx_mult_real_q15.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_cmplx_mult_real_q15.c
00004  * Description:  Q15 complex by real multiplication
00005  *
00006  * $Date:        27. January 2017
00007  * $Revision:    V.1.5.1
00008  *
00009  * Target Processor: Cortex-M cores
00010  * -------------------------------------------------------------------- */
00011 /*
00012  * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved.
00013  *
00014  * SPDX-License-Identifier: Apache-2.0
00015  *
00016  * Licensed under the Apache License, Version 2.0 (the License); you may
00017  * not use this file except in compliance with the License.
00018  * You may obtain a copy of the License at
00019  *
00020  * www.apache.org/licenses/LICENSE-2.0
00021  *
00022  * Unless required by applicable law or agreed to in writing, software
00023  * distributed under the License is distributed on an AS IS BASIS, WITHOUT
00024  * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
00025  * See the License for the specific language governing permissions and
00026  * limitations under the License.
00027  */
00028 
00029 #include "arm_math.h"
00030 
00031 /**
00032  * @ingroup groupCmplxMath
00033  */
00034 
00035 /**
00036  * @addtogroup CmplxByRealMult
00037  * @{
00038  */
00039 
00040 
00041 /**
00042  * @brief  Q15 complex-by-real multiplication
00043  * @param[in]  *pSrcCmplx points to the complex input vector
00044  * @param[in]  *pSrcReal points to the real input vector
00045  * @param[out]  *pCmplxDst points to the complex output vector
00046  * @param[in]  numSamples number of samples in each vector
00047  * @return none.
00048  *
00049  * <b>Scaling and Overflow Behavior:</b>
00050  * \par
00051  * The function uses saturating arithmetic.
00052  * Results outside of the allowable Q15 range [0x8000 0x7FFF] will be saturated.
00053  */
00054 
00055 void arm_cmplx_mult_real_q15(
00056   q15_t * pSrcCmplx,
00057   q15_t * pSrcReal,
00058   q15_t * pCmplxDst,
00059   uint32_t numSamples)
00060 {
00061   q15_t in;                                      /* Temporary variable to store input value */
00062 
00063 #if defined (ARM_MATH_DSP)
00064 
00065   /* Run the below code for Cortex-M4 and Cortex-M3 */
00066   uint32_t blkCnt;                               /* loop counters */
00067   q31_t inA1, inA2;                              /* Temporary variables to hold input data */
00068   q31_t inB1;                                    /* Temporary variables to hold input data */
00069   q15_t out1, out2, out3, out4;                  /* Temporary variables to hold output data */
00070   q31_t mul1, mul2, mul3, mul4;                  /* Temporary variables to hold intermediate data */
00071 
00072   /* loop Unrolling */
00073   blkCnt = numSamples >> 2U;
00074 
00075   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00076    ** a second loop below computes the remaining 1 to 3 samples. */
00077   while (blkCnt > 0U)
00078   {
00079     /* C[2 * i] = A[2 * i] * B[i].            */
00080     /* C[2 * i + 1] = A[2 * i + 1] * B[i].        */
00081     /* read complex number both real and imaginary from complex input buffer */
00082     inA1 = *__SIMD32(pSrcCmplx)++;
00083     /* read two real values at a time from real input buffer */
00084     inB1 = *__SIMD32(pSrcReal)++;
00085     /* read complex number both real and imaginary from complex input buffer */
00086     inA2 = *__SIMD32(pSrcCmplx)++;
00087 
00088     /* multiply complex number with real numbers */
00089 #ifndef ARM_MATH_BIG_ENDIAN
00090 
00091     mul1 = (q31_t) ((q15_t) (inA1) * (q15_t) (inB1));
00092     mul2 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1));
00093     mul3 = (q31_t) ((q15_t) (inA2) * (q15_t) (inB1 >> 16));
00094     mul4 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) (inB1 >> 16));
00095 
00096 #else
00097 
00098     mul2 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1 >> 16));
00099     mul1 = (q31_t) ((q15_t) inA1 * (q15_t) (inB1 >> 16));
00100     mul4 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) inB1);
00101     mul3 = (q31_t) ((q15_t) inA2 * (q15_t) inB1);
00102 
00103 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
00104 
00105     /* saturate the result */
00106     out1 = (q15_t) __SSAT(mul1 >> 15U, 16);
00107     out2 = (q15_t) __SSAT(mul2 >> 15U, 16);
00108     out3 = (q15_t) __SSAT(mul3 >> 15U, 16);
00109     out4 = (q15_t) __SSAT(mul4 >> 15U, 16);
00110 
00111     /* pack real and imaginary outputs and store them to destination */
00112     *__SIMD32(pCmplxDst)++ = __PKHBT(out1, out2, 16);
00113     *__SIMD32(pCmplxDst)++ = __PKHBT(out3, out4, 16);
00114 
00115     inA1 = *__SIMD32(pSrcCmplx)++;
00116     inB1 = *__SIMD32(pSrcReal)++;
00117     inA2 = *__SIMD32(pSrcCmplx)++;
00118 
00119 #ifndef ARM_MATH_BIG_ENDIAN
00120 
00121     mul1 = (q31_t) ((q15_t) (inA1) * (q15_t) (inB1));
00122     mul2 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1));
00123     mul3 = (q31_t) ((q15_t) (inA2) * (q15_t) (inB1 >> 16));
00124     mul4 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) (inB1 >> 16));
00125 
00126 #else
00127 
00128     mul2 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1 >> 16));
00129     mul1 = (q31_t) ((q15_t) inA1 * (q15_t) (inB1 >> 16));
00130     mul4 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) inB1);
00131     mul3 = (q31_t) ((q15_t) inA2 * (q15_t) inB1);
00132 
00133 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
00134 
00135     out1 = (q15_t) __SSAT(mul1 >> 15U, 16);
00136     out2 = (q15_t) __SSAT(mul2 >> 15U, 16);
00137     out3 = (q15_t) __SSAT(mul3 >> 15U, 16);
00138     out4 = (q15_t) __SSAT(mul4 >> 15U, 16);
00139 
00140     *__SIMD32(pCmplxDst)++ = __PKHBT(out1, out2, 16);
00141     *__SIMD32(pCmplxDst)++ = __PKHBT(out3, out4, 16);
00142 
00143     /* Decrement the numSamples loop counter */
00144     blkCnt--;
00145   }
00146 
00147   /* If the numSamples is not a multiple of 4, compute any remaining output samples here.
00148    ** No loop unrolling is used. */
00149   blkCnt = numSamples % 0x4U;
00150 
00151   while (blkCnt > 0U)
00152   {
00153     /* C[2 * i] = A[2 * i] * B[i].            */
00154     /* C[2 * i + 1] = A[2 * i + 1] * B[i].        */
00155     in = *pSrcReal++;
00156     /* store the result in the destination buffer. */
00157     *pCmplxDst++ =
00158       (q15_t) __SSAT((((q31_t) (*pSrcCmplx++) * (in)) >> 15), 16);
00159     *pCmplxDst++ =
00160       (q15_t) __SSAT((((q31_t) (*pSrcCmplx++) * (in)) >> 15), 16);
00161 
00162     /* Decrement the numSamples loop counter */
00163     blkCnt--;
00164   }
00165 
00166 #else
00167 
00168   /* Run the below code for Cortex-M0 */
00169 
00170   while (numSamples > 0U)
00171   {
00172     /* realOut = realA * realB.            */
00173     /* imagOut = imagA * realB.                */
00174     in = *pSrcReal++;
00175     /* store the result in the destination buffer. */
00176     *pCmplxDst++ =
00177       (q15_t) __SSAT((((q31_t) (*pSrcCmplx++) * (in)) >> 15), 16);
00178     *pCmplxDst++ =
00179       (q15_t) __SSAT((((q31_t) (*pSrcCmplx++) * (in)) >> 15), 16);
00180 
00181     /* Decrement the numSamples loop counter */
00182     numSamples--;
00183   }
00184 
00185 #endif /* #if defined (ARM_MATH_DSP) */
00186 
00187 }
00188 
00189 /**
00190  * @} end of CmplxByRealMult group
00191  */
00192